Start with the work you want to improve.
You may have a process that takes too long, gets rebuilt each time, or relies on a few experienced people. Its knowledge may sit across documents, software, conversations, and individual judgment. That is enough to begin.
I look at how the work happens, where it slows, what requires judgment, and what a useful outcome looks like. When an agent fits the process, I can build it.
What the right AI agent should handle
A useful agent needs a defined job and what it requires to perform well: approved knowledge, connected tools, a repeatable process, and human judgment or approval where needed.
I start with the outcome, then define the right capability, access, guardrails, and human oversight.
Built from real AI work
I’ve spent the past three years building with and for AI: Human dnAI, lead enrichment and outreach, research and market intelligence, managed sites, referral infrastructure, and installable tools for agent builders.
That work helps me identify where an agent adds value, what must be structured around it, and what reliable execution requires.
What your agent may need
The agent is one part of the system. Its usefulness depends on the business context, instructions, tools, and workflow built around it.
Knowledge
Your agent needs reliable context: products, policies, research, brand, methods, and past decisions. I identify what it should know, which sources to trust, and how that knowledge guides its work.
Skills
Skills define specific jobs: research a company, qualify a lead, prepare outreach, create a report, refresh content, or complete a process unique to your business. Each has a clear purpose and output.
Tools
Tools let the agent move work forward through research, internal data, email, spreadsheets, content systems, business software, or custom integrations. I define its access and keep selected actions under human control.
Workflows
Workflows create a repeatable path from request to result: steps, decisions, tools, approvals, exceptions, and outputs. The agent is built around the real process your business uses to create value.
A practical example: lead enrichment for agencies
Agency prospecting can consume hours across search, cleanup, research, qualification, outreach preparation, and follow-up. I built a lead enrichment system that connects those steps in one repeatable workflow.
It finds agency leads through repeatable LinkedIn Sales Navigator filters, scores activity signals, removes weak records, and sends stronger leads into deeper research. That context produces message hooks, offer angles, subject lines, and outreach, with Google Sheets sync and follow-up tracking.
The team gets hours back, applies one qualification standard, and identifies stronger opportunities before investing in deeper research.
Built for agents that use tools and take action
I also build tools agents can discover, install, configure, and use in real projects, including Transmitter’s Rewards packages.
@transmitter/referral-tracker
@transmitter/rewards-engine
In Cursor, Claude Code, and OpenClaw, an agent can discover a package, run checks, scaffold routes, configure the project, and embed referral information. Your agent may need to research, create, update, monitor, communicate, or coordinate. I define the job and build the tools it requires.
Process
-
(Work)
Name the job worth improving
Bring a defined agent idea, repeatable process, or bottleneck. I work with you to define what the work must accomplish and what a better result looks like.
-
(Fit)
Decide if an agent is the answer
I define the required knowledge, tools, workflow, guardrails, and human approvals. If another approach fits better, you still leave with a clear diagnosis.
-
(Run)
Put it on real work
Build around the real process, prove it on actual tasks, then extend it. The first agent must make valuable work easier to complete, improve, and scale.
Bring me the work you want to improve.
I’ll help you find the right answer and, when an agent fits, take it from idea to working system.
Book AI Strategy Session · $250